INTERNATIONAL JOURNAL OF HUMAN–COMPUTER INTERACTION, 20(1), 19–33

Copyright © 2006, Lawrence Erlbaum Associates, Inc.

Interruption Management and

Telephone Call Screening

Allen E. Milewski
Department of Software Engineering, Monmouth University

Notification-oriented computer interfaces are growing in importance, as is our under-

standing of how users manage interruptions. To gain insights relevant to the design of
such interfaces, this study explored telephone call screening as a common example of
how people manage technology-driven interruptions in their everyday residential en-
vironment. Survey results showed that audio screening is a frequent and regular prac-
tice, often used as part of an active, multicue strategy for managing interruptions. Fre-
quent screeners estimated knowing more about the call before engaging in social
interaction and were more selective when answering. Screening was viewed as an ef-
fective means of managing interruptions because screened, unanswered calls were
rated as less disruptive of ongoing activities than calls that were answered. These find-
ings on how people manage telephone call interruptions provide insights for design-
ers of notification systems. First, audio alerts can be highly effective, especially in resi-
dential settings. Second, alerts need not be simple to be effective; they can usefully
include rich information that clarifies how to respond to the interruption. Finally, in-
formative alerts offer a situated alternative to “presence publishing” interfaces.

1. INTRODUCTION

In the field of human–computer interaction, there is currently great interest in the

design of notifications and alerts and in the consequences of interrupting users
while they are performing tasks (Horvitz, Koch, & Apacible, 2004). The motivation
for this interest includes the growing prevalence of synchronous communication
systems, especially instant messaging (IM) and alerting services, such as Microsoft
.NET™ Alerts. Several researchers have argued that these notifications and inter-
ruptions require a switch in focus for user interface designers, from interfaces for
user-initiated and controlled activities toward a new kind of interface where users
must integrate their own ongoing activities with those that are computer initiated
(e.g., McCrickard & Chewar, 2003).

Correspondence should be addressed to Allen E. Milewski, Monmouth University, Cedar Avenue,

West Long Branch, NJ 07764–1898. E-mail: amilewsk@monmouth.edu
20 Milewski

Recent research has suggested that careful design of interrupting notifications

can determine their effectiveness for computer interfaces and in other domains
such as cockpit and flight deck environments (McFarlane & Latorella, 2002). Users’
performances on ongoing tasks are often disrupted by incoming notification
(Adamczyk & Bailey, 2004; Cutrell, Czerwinski, & Horvitz, 2000), but how effi-
ciently and effectively the interruption is handled by users depends on characteris-
tics of the notification as well as task characteristics. For example, Latorella (1999)
found that individuals initiate the handling of auditory interruptions more quickly
than visual interruptions and that auditory interruptions were more disruptive to
ongoing performance, although interruption modality effects also depend on task
modality in complex ways. Notification timing and response requirements are also
important factors. For example, interruptions are less disruptive if delayed until
intertask periods or natural break points (Adamczyk & Bailey, 2004; Czerwinski,
Cutrell, & Horvitz, 2000). Finally, a key characteristic determining the disruptive
effects of alerts has to do with response requirements. McFarlane (1999) delineated
four situations: (a) Immediate interruption requires the recipient to respond without
delay, (b) negotiated interruption involves an announcement that an interruption is
needed and then lets its recipient negotiate when and how the interruption will be
dealt with, (c) mediated interruption allows for a third party (e.g., an agent) to receive
the interruption and plan a response, and (d) scheduled interruption restricts inter-
ruptions to prearranged times. McFarlane found that negotiated interruptions re-
sulted in the best performance except when extreme timeliness was critical.
Much of the past research on interruption has been done in a carefully con-
trolled, dual-task paradigm where participants are engaged in one specific, ongo-
ing task and the demand characteristics for responding to a notification are high.
Moreover, the majority of recent studies have concentrated on visual alerts in a
desktop computing context. This research has yielded important insights about in-
terruptions at a microlevel of memory load and basic cognitive processing.
McCrickard, Czerwinski, and Bartram (2003) suggested that the investigation of
notification management must broaden to include “as much of the human infor-
mation awareness need domain as possible” (p. 513). Broadening includes moving
beyond the computer desktop but also could include a fuller understanding of how
the basic processes are utilized in a wider range of everyday environments. In ev-
eryday, natural settings, it is likely that an additional factor influencing users’ reac-
tions to interruption includes the goal-oriented strategies users adopt to deal with
distraction (Latorella, 1999; McCrickard & Chewar, 2003). For example,
Czerwinski, Horvitz, and Wilhite (2004), in a diary study of office work, reported
frequent and deliberate task-switching activities. Residential interviews and
self-reports by Nagel, Hudson, and Abowd (2004) revealed that participants’ will-
ingness to handle interruptions varied across individuals with their current loca-
tion in the home as well as with their current activity.
Over the past several decades, one of the most prominent scenarios for interrup-
tion is clearly the telephone call. The average business day is filled with a signifi-
cant number of telephone-related interruptions (Czerwinski et al., 2004; O’Conaill
& Frohlich, 1995). Residential telephone users are also often inundated with calls
(Lacohée & Anderson, 2001), and in a series of studies of residential telephony us-
Interruption Management and Call Screening 21

age, Gillard, Bow, and Wale (1995) determined that the single most critical issue for
residential telephone users is “access management,” which includes protecting
oneself from constant interruption by unwanted calls.
In its simplest form, the telephone call is far from a perfect notification. The ring-
ing of the telephone signifies that some potentially important event is occurring—
namely, that someone wants to communicate. To determine the true importance of
the event compared with ongoing activities, the recipient of the notification has to an-
swer the call directly and engage in social interaction. Doing so can be taxing in terms
of social effort and has a large impact on whatever behavior was ongoing when the
interruption occurred. Those that do not respond to the notification may have no
way of ever determining whether they should have responded. Aperson engaged in
some critically important task when the phone rings is faced with this dilemma, and
common experience shows that people generally respond to this notification.
The standard telephone ring, because it demands a high-overhead response,
could be classified as an immediate interruption according to McFarlane’s (1999)
taxonomy, and this type of interruption has significant disruptive effects on the
performance of ongoing activities. In the telephony domain, there are alternative
methods for handling incoming calls—audio call screening is one of them (Gillard
et al., 1995; Lacohée & Anderson, 2001). Audio call screening is more similar to
McFarlane’s negotiated interruption, which is superior to other types of interrup-
tion in that it causes less disruption. With audio call screening, the caller indicates
that a conversation is desired by adding verbal information about the incoming call
that can help the recipient determine its priority without engaging in social interac-
tion. Generally, this audio information is played out loud while being captured as a
message with a telephone answering machine.
The similarities of audio call screening with negotiated interruption are signifi-
cant. First, no immediate response is required when audio screening, other than lis-
tening to the message. Indeed, audio screening does not even require close proxim-
ity to the telephone. It can be argued that listening to the audio message entails
cognitive overhead in itself, but the relatively low overhead of listening can be used
to manage the much higher overhead of social interaction required to engage in the
call. Trade-offs between the disruptive aspects of interruptions and their critical in-
formation value have been recognized in both theory and empirical findings on in-
terruptions (McCrickard, Chewar, Somervell, & Ndiwalana, 2003). Second, the re-
cipient of a screened call has control over if and when the interruption is handled,
and this control is situational (Suchman, 1987). With no prior planning and based
on the immediate context, users can decide whether the alert is important enough
to allow further interruption by responding to it.
Finally, in the case of audio screening, negotiation of the interruption is indirect
and asynchronous. The recipient’s side of the negotiation consists of deciding to
answer the call or not. The caller’s side consists of providing a potentially rich array
of information describing the nature of the call. During audio screening, for exam-
ple, a caller is likely to indicate his or her identity and the reason for calling. An in-
dication of the caller’s own future availability might also be included. This direct
information is only part of the message, however. The voice quality of a verbal mes-
sage often indicates its emotional content (Yang, 1996), so that recipients may de-
22 Milewski

tect, for example, its friendliness versus tension. Finally, callers may even indirectly
infer both urgency and call length. All of these cues provide information useful in
deciding whether to answer the call. If one were to create a continuum of informa-
tion richness, audio screening would be an example of highly “informative alert-
ing” compared with other alerts such as the telephone ring. It is this informative
alerting that gives recipients the control to manage further prolonged interruption
with only the small overhead of listening. Therefore, informative alerting is one of
the key aspects of similarity between audio screening and negotiated interruption.
Given the similarity to the highly efficient negotiated interruptions, one might
expect that call screening would reduce the perceived disruption of telephone calls
and be used frequently as part of telephone users’ interruption management strate-
gies. McFarlane’s (1999) categorization of interruption types was initially intended
only as a framework of system design alternatives, and there has been little study
of how these interruption types interact with the strategies users develop for deal-
ing with interruptions. Nonetheless, one might expect that, given the choice, users
would incorporate more efficient interruption types into their activities and that
audio screening would be an example. However, beyond the basic finding that tele-
phone users do screen calls (Gillard et al., 1995; Lacohée & Anderson, 2001), little is
known about it. For example, there is no normative data indicating how often peo-
ple screen calls or what information they concentrate on when they do. It is also not
clear how disruptive telephone calls are to daily activities and to what degree audio
screening influences the disruption. Similarly, little is known about the communi-
cations aspect of screened calls: How do recipients use information to negotiate
conversations, and what are the social implications of call screening?
The exploratory study of audio call screening presented here has the goal of
better understanding how informative alerting is used to manage interruptions
and negotiate communications during everyday activities. The study investigates
the frequency, motivations, and basic characteristics of audio call screening. In ad-
dition, because the particular interruptions handled by call screening are of a
strongly social nature, this study also explores the social acceptability of screening.

2. METHOD

A Web-based questionnaire was sent to 700 employees of a large telecommunica-

tions company. Employees were chosen randomly from the corporate directory
with the restrictions that they be employed within the United States. An e-mail in-
vitation describing the survey was accompanied by a link to the online survey. The
only reward for responding was the chance of winning a drawing for one $35.00
gift certificate; the response rate was 27%. The 187 respondents were distributed
among technical, marketing, and administrative staff positions. There were 109
women and 78 men. Mean age was 43.2 years (SD = 9.3 years).
The survey consisted of 26 items and was estimated to take approximately 5 to
10 min to fill out. Survey items were developed to obtain direct feedback from re-
spondents about the degree to which audio screening is used as a regular and effec-
tive interruption management technique. The issues these items probed included
Interruption Management and Call Screening 23

the frequency of screening, reasons for screening, the information utilized while
screening, and the disruptiveness of answered telephone calls and audio-screened
telephone calls. Additional survey items probed the social acceptability of audio
screening and recipient demographics. Some survey items had constrained re-
sponses and were automatically compiled; other items allowed open-ended re-
sponses. Not every item was completed by every respondent.

3. RESULTS AND DISCUSSION

Survey results were analyzed and organized around the following key questions
regarding call-screening behavior.

3.1. How Frequently Do Telephone Users Auditorily Screen Calls?

Of 187 total respondents, 147 (79%) reported having an answering machine. The
preponderance of answering machines was in a residential setting. Of those with
answering machines, 83% were situated in residences, whereas 15% were in of-
fices. Three percent of the respondents reported having answering machines in
both locations.
Of the respondents with answering machines, 104 (71%) reported engaging in
audio screening for at least some calls, with about one third screening more than
half their calls. Across all screeners, the mean estimated percentage of calls
screened was 46.2%, but there was a good deal of variation between respondents in
estimated screening frequency. Indeed, the distribution of audio-screening fre-
quency, shown in Figure 1, was strongly bimodal, suggesting two classes of respon-

FIGURE 1 Screening frequency.

24 Milewski

dents differing in their screening behavior: frequent screeners and infrequent

screeners. For subsequent analyses, these two groups were divided according to a
median split across the “% calls screened” dimension (Mdn = 33).
The two groups differing in estimated screening frequency also differed in sev-
eral other measures, suggesting different interruption management strategies (see
next) but were not found to differ in demographics (i.e., age, gender).

3.2. Why Do Telephone Users Screen?

Virtually all of the respondents that do screen (92%) report audio screening spe-
cifically to decide whether to answer the call. This, for example, compared with
simply capturing the caller’s message or as a way to fill time while getting to the
telephone.
The fact that screening is part of an active interruption management strategy is
bolstered by differences between frequent and infrequent screeners. Specifically,
frequent screeners estimated not answering a larger percentage of calls because of
something heard in the screened message. Overall, rejection rates were correlated
with the percentage of calls screened, r(98) = .22, p < .05, and the two groups dif-
fered significantly, t(98) = 2.19, p < .05. The rejected messages were still recorded as
voice messages, but frequent screeners were more selective in deciding to engage
in interactive conversation. Moreover, frequent screeners had a higher estimated
percentage of knowing, before answering, who the caller was, t(96) = 2.149, p < .05,
and what the call was about, t(87) = 1.74, p < .075 (marginally significant ).
Ninety-seven percent of the respondents reported that receiving unwanted tele-
phone calls was an intrusion and disrupted their ongoing activities. Thirty-nine
percent reported “hating it,” 23% categorized unwanted calls as “a lot of trouble,”
and 35% said they were a minor inconvenience. Only 3% viewed unwanted calls as
“no trouble at all.” There were no differences between frequent and infrequent
screeners in their judgment of unwanted calls. The perceived intrusiveness of un-
wanted calls varied with the amount of work they require, possibly because more
answering activity interrupts and interferes more with ongoing tasks. Participants
were asked about the intrusiveness of unwanted calls when they answered them
versus when they did not answer them. The distributions of responses about intru-
siveness were significantly different, with the intrusiveness of unwanted but unan-
swered calls much lower, χ2(3, 186) = 53.2, p < .001. For example, 20% of respon-
dents viewed these to be “no trouble at all,” and 52% considered them to be only a
minor inconvenience. Finally, when asked about the intrusiveness of listening to
audio screening, 60% indicated that listening to audio screening was an intrusion,
whereas 40% viewed this as not at all intrusive or disruptive. Frequent screeners
judged audio screening to be less intrusive and disruptive than did infrequent
screeners, t(102) = 2.73, p < .01.
In summary, telephone users appear to screen because it is a relatively easy and
effective way to manage interruptions. The prevalence of unwanted calls and the
major concern with finding effective ways of managing them is consistent with
past findings that access management is a major residential concern (Gillard et al.,
Interruption Management and Call Screening 25

1995; Lacohée & Anderson, 2001). Moreover, the perception that screened calls are
less disruptive than answered calls is consistent with the view that audio screening
shares important characteristics with McFarlane’s (1999) negotiated interruptions.
McFarlane found better performance and less disruption when participants were
given negotiated compared with immediate interruptions. In the study presented
here, respondents have a choice of immediately answering the telephone or using
audio screening to decide whether to answer. The fact that a significant proportion
of telephone users choose to screen suggests that they have incorporated the effi-
ciency of negotiated interruptions into their interruption management strategies.
Moreover, the observation that those who perceive audio screening to be less dis-
ruptive also screen more may further suggest a deliberate strategy.

3.3. What Information Is Used During Screening?

Respondents were asked to rate, on a 7-point scale, the importance of the several
categories of information they might obtain while screening. These categories were
developed prior to the survey during informal interviews conducted with several
colleagues. These colleagues were asked to speculate about what information
might be available in an audio screening. They are caller’s name, reason for call,
caller’s telephone number, urgency of call, length of call, recipient of the call, and
emotional impact. Ratings ranged from 7 (almost always important) to 1 (rarely impor-
tant). Rating results are shown in Figure 2 averaged across all respondents.
There was a strong effect of information category, F(6, 101) = 91.6, p < .01. The re-
sults of Wilcoxon pairwise tests reveal several differences in rated importance.
Caller’s name or identity was most important. It was rated significantly higher in
importance than each of the others. Urgency of the call was next, rated significantly

FIGURE 2 Seven-point importance rating for information obtained while screening.

26 Milewski

more important than either reason for the call (z = 3.98, p < .01) or emotional impact
(z = 4.19, p < .01), although still less important than the caller’s identity (z = 4.87, p <
.01). There were no statistical differences between reason and emotion. The recipi-
ent of the call1 was rated next in importance; its rating was less that either reason (z
= 4.16, p < .01) or emotion (z = 4.05, p < .01) but higher than caller’s telephone num-
ber (z = 3.26, p < .01) and length of call (z = 4.02, p < .01).
The caller’s telephone number and the estimated length of the call were rated
least important.
Rating data were similar to qualitative answers about what it is that makes users
pick up a call or not, after screening. Seventy-two percent of respondents (73 of 102)
mentioned the caller’s identity only. Of these, 5 respondents described making a
simple determination of whether the caller was a telemarketer. Six percent men-
tioned only the reason for the call, and 14% mentioned both caller’s identity and
reason for calling for deciding to answer. This means 20% were concerned with the
reason for the call as a factor in answering calls. A substantial number of responses
also indicated that their behavior depended on call context, such as the time of day,
ongoing activities, and so on.
Given the importance of context and the primacy of callers’ identity, another
way of determining the importance of other information is to ask “Once you have
heard who is calling, how often do you wait to hear the reason for the call?” To this
question, only 12% said they never waited to hear the reason for the call before
picking up; 88% apparently used the reason for the call at least occasionally in de-
ciding to answer (Figure 3). Although only 3% reported always waiting, 24% re-
ported usually waiting to hear the reason of the call before picking up. Sixty-two
percent seldom waited.
Frequent screeners and infrequent screeners had very similar information-gath-
ering patterns. Rankings of importance across information type did not differ for
the two groups, and the likelihood of waiting for the reason for the call after hear-
ing the caller’s identity was the same for frequent screeners as for infrequent
screeners.
Across all screeners, information about caller identity and reason for the call
were estimated to be detected quite accurately, although not perfectly. Respon-
dents indicated that, after screening, they had a good sense of who was calling a
mean of 79.2% (SD = 27.3) of the time (Mdn = 90%) and, similarly, that they could
predict the reason for the call a mean of 75.3% (SD = 29.0) of the time (Mdn = 90%).
One of the reasons these predictions were not higher was that, according to several
respondents, telemarketers often purposely do not make their identity clear until
after a human answers the call.
In summary, caller identity appears to be the most critical piece of information
gathered while screening by far. However, the reason for the call as well as other
contextual factors, such as emotional impact, urgency, and so forth, is also impor-
tant. This pattern of behavior for telephony conversations is much like the begin-
ning of conversations in a face-to-face setting and is consistent with the notion that

1Information about the call recipient could be useful in a household situation where there are several

potential recipients sharing a line.

Interruption Management and Call Screening 27

FIGURE 3 Estimated likeli-

hood of waiting to hear reason
for call before answering.

conversation “starts” on the telephone are negotiated in a way that is analogous to

when one is interrupted by someone who desires a face-to-face conversation
(Clark, 1996).

3.4. How Do Telephone Users Screen?

Those respondents that did screen reported having their answering machines set to
an average of 3.92 rings. The majority of respondents had machines set to 4 rings,
but 11 (of 104) had them set to 2 rings. These respondents were evenly distributed
across the range of proportion of calls screening, making it unlikely that these re-
spondents consciously had set their equipment to facilitate screening. Indeed, the
number of rings was not a reliable predictor of any screening behavior. Frequent
and infrequent screeners did not differ in this measure, and even those that re-
ported screening 100% of their calls were evenly distributed across the range of
ring settings. It is not clear if this similarity across groups reflects a failure to con-
sciously plan for screening or simply not knowing how to modify ring settings in
what is often an obscure answering machine user interface.
The survey question “How long do you screen” was open ended and could be
answered with either a number or a text phrase. Seventy-two percent of respon-
dents answered with a numeric duration estimate, and the mean estimate was 4.46
sec (Mdn = 3 sec). There was a small difference in listening duration means between
frequent screeners (5.1 sec) and infrequent screeners (3.7 sec), but the difference
was not significant and resulted from very long durations reported by two frequent
screeners.
Survey responses indicated an interesting interaction between audio screening
and screening with caller ID displays. Fifty-seven percent of all respondents re-
ported having and using caller ID display devices. Those respondents with caller
ID devices were no less likely to engage in audio call screening than those without
caller ID devices. In fact, those respondents with caller ID service reported audio
screening a significantly larger percentage of messages (48.5%) than those respon-
dents without caller ID service (29.6%), t(94) = 2.82, p < .01. Consistent with this, re-
28 Milewski

spondents with caller ID service were also more likely to be categorized as frequent
screeners.
The estimated number of calls screened was the same for auditory screening
(38.6%) and screening with caller ID service (41.3%). One interpretation is that re-
spondents with caller ID service and answering machines used both kinds of infor-
mation. It is not clear from the data if these are used in parallel or by some other
strategy. In open-ended answers, five respondents explicitly said they engaged in
audio screening only if the caller ID display failed to show a number. Respondents
with caller ID rated the visual caller ID to be significantly less intrusive than audio
screening, t(103) = 6.4, p < .01. However, several respondents noted that a signifi-
cant advantage for audio screening in the residential setting was that it could be
conveniently used from any room in the house.

3.5. How Is Call Screening Viewed?

Screeners were asked to compare their own likelihood of screening incoming calls
with the likelihood that calls they initiate are screened by others. Both frequent and
infrequent screeners estimated the likelihood of being screened as less than their
own screening frequency (Figure 4). The overall estimate of screeners being
screened was 28.5% (i.e., frequent and infrequent screeners averaged together).
This was a relatively accurate reflection of the likelihood of actually being screened.
If all respondents with answering machines are taken into account regardless of
whether they screen, the total screening frequency was 32.6%. Frequent screeners
tended to overestimate screening, whereas infrequent screeners underestimated
screening. In summary, although participants’ own screening behavior slightly bi-
ased their perceptions of how often other people screen, their estimates generally
matched reality.

FIGURE 4 Estimated percentage of screening and being screened.

Interruption Management and Call Screening 29

The social acceptability of screening has often been questioned (Martin, 1996).
However, out of all respondents that screen, only one respondent considered their
own screening to be “always rude.” Seventy-six percent considered their own
screening to be “never rude,” and the remaining 23% responded that it depended
on whom they were screening. There were no significant differences between fre-
quent and infrequent screeners.
Similarly, when all respondents were asked if it was rude for other people to
screen them during a call that they initiated, 79% responded that it was not rude,
whereas 21% thought it was. There were no differences between respondents who
screened and those who did not screen.
Despite strong agreement among screeners that screening is not rude, screeners
virtually never revealed to people calling them that their calls might be screened.
Only 3 of 104 screeners reported indicating in their answering machine greeting
that the call might be screened. This tendency to leave the situation ambiguous
may indicate some doubt over screening’s social acceptability, although it could
also indicate a courtesy to the caller. Adding this warning to a greeting increases its
duration and complexity, as it might or might not be true for any specific call.

4. CONCLUSIONS

Notification-based interfaces and interruption management has received a great

deal of attention in the office desktop environment. This study set out to under-
stand telephone screening with the intention of drawing insights about informa-
tive alerting and extending our appreciation of interruption management to the
residential environment. The survey data clearly showed that unwanted telephone
calls are a common occurrence and that they pose a serious inconvenience and dis-
ruption of recipients’ ongoing tasks. Just as Gillard et al. (1995) reported, access
management is a major household concern, and just as in the workplace (e.g.,
Czerwinski et al., 2004), there is a need for interruption-management tools coupled
with a useful strategy. In the home, the residential answering machine and audio
call screening are common elements of that strategy. Audio screening is a common
practice that is performed regularly by a significant subset of telephone users.
These data show that audio screening is not considered to be as disruptive as an
uninformative call alert (e.g., “ring”) that requires social interaction to respond to.
This is consistent with theoretical models of interruption and notification
(McFarlane, 1999), which show that the least disruptive alerts are those that do not
require an immediate response, provide the recipient with a degree of control over
when they are handled and therefore have minimal effect on ongoing tasks. Audio
screening has these characteristics in common with McFarlane’s negotiated inter-
ruptions, and it is interesting to note that in both cases there is cognitive overhead
associated with the interruption. It has been shown that the cognitive overhead as-
sociated with interruptions can result in degraded performance in a variety of tasks
(Cutrell et al., 2000; Czerwinski et al., 2000). However, the trade-off of an interrup-
tion’s cognitive overhead compared with its information value is critical
30 Milewski

(McCrickard et al., 2003), and the finding that audio screening is a common activity
suggests that its value is high.
An interesting suggestion from the findings presented here is that screening
may be part of an active, deliberate strategy for negotiating interruptions. This is
evidenced by the fact that those who screen more are more selective in answering
calls, feel they know more about the call prior to interacting with the caller, and
judge audio screening as less disruptive. Indeed, audio screening may often be just
part of an interruption-management strategy that can also include caller ID tech-
nology because their use overlaps.
It has recently been shown (Hudson, Christensen, Kellog, & Erickson, 2002) that
corporate research managers have a complex relationship with communications
that interrupt workday activities. On one hand, these interruptions are viewed as
an irritation that decreases the efficiency of ongoing tasks, but at the same time they
are understood to be necessary in order to keep up with new information. In other
words, users are aware of the trade-offs associated with interruptions. These find-
ings are consistent with and extend this view. Several respondents indicated that
the cognitive overhead associated with interrupting ongoing activities to screen a
call was preferable to the alternative of not screening and either (a) missing an im-
portant call or (b) answering the call and risking being forced into social interaction
of potentially much greater overhead. It appears that people in both management
and residential settings understand the value of these interruptions and actively
develop strategies to make use of them.
Compared with many kinds of computer-based notifications, telephone call
screening provides a much richer stimulus. The rich set of information that is deliv-
ered directly from a speaking human can potentially be used to analyze the impor-
tance of the notification along dimensions of content, emotional impact, and ur-
gency. The data presented here suggest that when telephone users screen calls, they
do, in fact, utilize several of the potential cues. The identity of the caller is rated as
the most critical, but respondents report utilizing other characteristics of the call as
the situation requires. The reason for the call doubtlessly comes from the verbal
content, whereas the urgency of the call and emotional impact may be extracted
from other vocal characteristics. The caller’s telephone number was rated as sur-
prisingly unimportant compared with other cues, but this may be due to the joint
use of caller ID services along with screening as part of many interruption manage-
ment strategies. However, the general conclusion is that users utilize a wide array
of information in deciding whether to react to a notification.
Although these findings suggest that audio screening represents an effective
way to negotiate communications, there are aspects that could prevent attitudes to-
ward it from being universally positive. In this paradigm, the initiators of commu-
nication (i.e., callers) are required to provide information that is filtered by recipi-
ents. It is unlikely that negative attitudes in this case are the result of a work–benefit
mismatch between caller and recipient (Grudin, 1987), because the work involved
in the caller leaving a message benefits the caller by increasing the chances of reach-
ing the recipient (either immediately or later). It is more likely that there are issues
of social control when the caller feels that his or her priority is being judged by the
recipient. In the study presented here, a small number of respondents indicated
Interruption Management and Call Screening 31

negative attitudes toward screening, but respondents generally claimed that

screening is not considered to be rude. The finding that few respondents admit to
callers that their calls may be screened could suggest either that screening may not
be viewed as entirely acceptable or that respondents are hesitant to impose on call-
ers with longer, more complex greetings. In either case, there may be a need to ex-
plore techniques for obtaining information about the call from the caller that are
streamlined and inoffensive.
These results from telephone call screening have practical implications for the
design of notification systems in general. First, they suggest that audio alerts can
be very useful. Not only can they support rich information, but they can be dis-
pensed with quickly because of their transient nature. Informative audio alerts
may be especially useful in residential settings because they can be heard at a
distance. This would suggest that even for personal computer (PC) applications,
carefully constructed auditory alerts may be effective notifications in the residen-
tial environment where users may not spend lengthy periods of time in front of a
visual display.
Second, the results suggest that alerts need not be simple to be effective. Some
studies suggest presenting simple “warnings” with little information prior to pre-
senting notification content (see McFarlane & Latorella, 2002). However, this con-
clusion may apply best to contexts with high memory requirements and has more
to say about the ordering of information than its amount. Indeed, these results sug-
gest that residential alerts can be particularly useful when they provide a signifi-
cant amount of information that clarifies for the user how best to respond to the
alert. In telephone call screening, this includes caller identification, purpose, emo-
tional content, and so on. In other applications, the appropriate information may
depend on other factors, but the point is that useful alerts need not be limited in in-
formation content. Consistent with earlier theory (McFarlane, 1999), alerts can be
efficient if the cognitive overhead required to process the alert is small compared
with that required to respond to it. Informative alerts may require more attention
than simple ones but are still attractive if they can prevent further work.
It should be noted that audio call screening itself can benefit from design im-
provements. Although the informative alert is effective, it is also time consuming to
wait for the telephone to ring several times, then to wait for the greeting, and then
to listen to the caller’s message. One solution would be to eliminate the telephone
ring altogether and replace it with the caller’s verbal message (Milewski & Smith,
2003). Doing so would increase efficiency but retain the value of the informative
alert.
A growing number of PC applications help users to manage interruptions by
continually publishing updated information about recipients’ presence and avail-
ability. In popular IM systems, for example, users can indicate “I am available” or
“I am away” or “I am busy” and can have that status published to those who may
want to communicate with them. Informative alerting, of the type involved in au-
dio call screening, can be thought of as an alternative of sorts to these presence dis-
plays. It can be argued that properly constructed notifications that have rich infor-
mation but low response overhead provide a very efficient means for negotiating
communication. It is important to note that the data presented here do not directly
32 Milewski

compare the effectiveness of informative alerting with the presence publishing of-
ten used in IM systems. Each paradigm is likely to have its own advantages. How-
ever, taken together with previous findings that availability status displays are of-
ten incorrect and ignored (Milewski & Smith, 2000), these data suggest that
informative alerting may be relatively more workable, at least for telephony. In-
deed, the extent to which presence/availability status displays work at all for IM
systems may be the result of a fairly unique characteristic of the desktop PC. That
is, people who are available for messaging on the PC are nearly always performing
some measurable keyboard activity, which can be used as a powerful presence cue.
This is not true for landline telephones and is definitely not true for mobile or ubiq-
uitous computing devices, which can remain powered on but dormant for hours
even though the owners may be eagerly awaiting communication. Moreover,
Nardi, Whittaker, and Bradner (2000) discovered that even when text IM users on
PCs had status information available, they still negotiated availability in their first
few message interactions. Because text-based interaction may be less socially de-
manding than voice, Nardi et al. suggested that augmenting telephony with early
negotiation via text messages may prove useful. The data presented here suggest
that many telephone users already use audio-screening technologies in this active
way and that informative alerting could form the basis for useful presence-man-
agement strategies.

REFERENCES

Adamczyk, P. D., & Bailey B. P. (2004). If not now, when?: The effects of interruption at differ-
ent moments within task execution. In Human Factors in Computing Systems: Proceedings of
CHI’04 (pp. 271–278). New York: ACM Press.
Clark, H. (1996). Using language. Cambridge, UK: Cambridge University Press.
Cutrell, E., Czerwinski, M., & Horvitz, E. (2000). Effects of instant messaging interruptions
on computing tasks. In Extended Abstracts of CHI ‘2000, Human Factors in Computing Sys-
tems (pp. 99–100). New York: ACM Press.
Czerwinski, M., Cutrell, E., & Horvitz, E. (2000). Instant messaging: Effects of relevance and
time. In S. Turner & P. Turner (Eds.), People and Computers XIV: Proceedings of HCI 2000
(Vol. 2, pp. 71–76). Swindon, UK: British Computer Society.
Czerwinski, M., Horvitz, E., & Wilhite, S. (2004). A diary study of task switching and inter-
ruptions. In Human Factors in Computing Systems Proceedings of CHI’04 (pp. 175–182). New
York: ACM Press.
Gillard, P., Bow, A., & Wale, K. (1995). Privacy and control: Social indicators of interest in fu-
ture telecommunications. Telecommunications Needs Research Group Technical Report. Mel-
bourne, Australia: Royal Melbourne Institute of Technology.
Grudin, J. (1987). Social evaluation of the user interface: Who does the work and who gets
the benefit. In H.-J. Bullinger & B. Shackel (Eds.), INTERACT 1987. IFIP Conference on Hu-
man-Computer Interaction (pp. 805–811). Stuttgart, Germany: Elsevier Science.
Horvitz, E., Koch, P., & Apacible, J. (2004). BusyBody: Creating and fielding personalized
models of the cost of interruption. In Proceedings of Meetings of Computer Supported Collabo-
rative Work (pp. 507–510). New York: ACM Press.
Hudson, J. M., Christensen, J., Kellog, W. A., & Erickson, T. (2002). “I’d be overwhelmed, but
it’s just one more thing to do”: Availability and interruption in research management. In
Interruption Management and Call Screening 33

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 97-102).
New York: ACM Press.
Lacohée, H., & Anderson, B. (2001). Interacting with the telephone. International Journal of
Human-Computer Studies, 54, 665–699.
Latorella, K. A. (1999). Investigating interruptions: Implications for flightdeck performance
(NASA/TM–1999–209707). Washington, DC: National Aviation and Space Administra-
tion.
Martin, J. (1996). Miss Manners’ basic training: Communication. New York: Crown.
McCrickard, D. S., & Chewar, C. M. (2003). User goals and attention costs. Communications of
ACM, 46(3), 67–72.
McCrickard, D. S., Chewar, C. M., Somervell, J. P., & Ndiwalana, A. (2003). A model for noti-
fication systems evaluation–Assessing user goals for multitasking activity. ACM Transac-
tions on Computer-Human Interaction, 10, 312–338.
McCrickard, D. S., Czerwinski, M., & Bartram, L. (2003). Introduction: Design and evalua-
tion of notification user interfaces. International Journal of Human-Computer Studies, 58,
509–514.
McFarlane, D. (1999). Coordinating the interruption of people. In Human Computer Interac-
tion–Proceedings of INTERACT’99 (pp. 295–303). Amsterdam: IOS Press.
McFarlane, D. C., & Latorella K. A. (2002). The scope and importance of human interruption
in human-computer interaction design. Human–Computer Interaction, 17, 1–61.
Milewski, A. E., & Smith, T. (2000). Providing presence cues to telephone users. In Proceed-
ings of the ACM Conference on Computer Supported Cooperative Work (pp. 89–96). New York:
ACM Press.
Milewski, A., & Smith, T. (2003). Telephone voice-ringing using a transmitted voice announcement
(U.S. Patent No. 6,519,326). Washington, DC: U.S. Patent and Trademark Office.
Nagel, K. S., Hudson, J. M., & Abowd, G. D. (2004). Predictors of availability in home life
context-mediated communication. In Proceedings of the ACM Conference on Computer Sup-
ported Cooperative Work (pp. 497–506). New York: ACM Press.
Nardi, B., Whittaker, S., & Bradner, E. (2000). Interaction and outeraction: Instant messaging
in action. In Proceedings of the ACM Conference on Computer Supported Cooperative Work (pp.
79–88). New York: ACM Press.
O’Conaill, B., & Frohlich, D. (1995). Timespace in the workplace: Dealing with interruptions.
In Proceedings of CHI ’95 (pp. 262–263). New York: ACM Press.
Suchman, L. (1987). Plans and situated actions: The problem of human-machine communication.
New York: Cambridge University Press.
Yang, L. (1996). Interruptions and intonation. In Proceedings of the International Conference on
Spoken Language Processing ’96 (pp. 1872–1875). Retreived from
http://www.isca-speech.org/archive/icslp_1996/i96_1872.html